Is it possible to send different data to MCP App vs Agent?

[rohithrangaraju-tomtom]

Pre-submission Checklist

• I have checked that this question would not be more appropriate as an issue in a specific repository
• I have searched existing discussions and documentation for answers

Question Category

• Protocol Specification
• SDK Usage
• Server Implementation
• General Implementation
• Documentation
• Other

Your Question

The Challenge

The same content from a tool result goes to both:

The MCP App (via app.ontoolresult)

The AI agent (as tool result in conversation)
For geospatial tools, this is problematic:

App needs: Full coordinate arrays (legs[].points[] for route polylines, boundary[] for polygons)
Agent needs: Just summaries (distance, duration, counts) - coordinate arrays waste tokens

Is there a way in the current MCP Apps/MCP architecture to:

Send full data to the App for visualization
Send trimmed data to the agent

[imjvdn]

Yes—the documented app-only tool pattern fits this:

1. Have the model-facing route tool return only the distance/duration/count summary and a routeId.
2. Register a separate get_route_geometry tool with _meta: { ui: { visibility: ["app"] } }. It accepts routeId and returns the full coordinates.
3. In the App, receive the ID from the original tool result, then fetch the geometry:

const result = await app.callServerTool({
  name: "get_route_geometry",
  arguments: { routeId },
});

Render that result locally without sending the coordinate arrays through updateModelContext() or sendMessage(). The visibility specification requires the host to omit app-only tools from the model's tool list. The patterns guide also includes chunked app-only responses for geometry too large for a host's response limit.

This uses an extra App-to-server call, but the model-facing response stays small.